[PR #17] [MERGED] fix: name a capacity stockout, and find Python instead of assuming it #16

New issue

Closed

opened 2026-10-31 22:02:41 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-31 22:02:41 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/Comfy-Org/comfy-qa-tools/pull/17
Author: @aliranjah
Created: 8/25/2026
Status: ✅ Merged
Merged: 8/25/2026
Merged by: @aliranjah

Base: main ← Head: fix/stockout-and-python

📝 Commits (1)

ff3f014 fix: name a capacity stockout, and find Python instead of assuming it

📊 Changes

6 files changed (+128 additions, -4 deletions)

View changed files

📝 comfy_qa/lifecycle.py (+25 -0)
📝 comfy_qa/provision.py (+31 -4)
📝 docs/troubleshooting.md (+11 -0)
📝 tests/test_docs.py (+2 -0)
📝 tests/test_lifecycle.py (+38 -0)
📝 tests/test_provision.py (+21 -0)

📄 Description

Two findings from a live run, one of which had been misleading us for an hour.

The box could not start — Google had no L4 capacity

The zone does not have enough resources available to fulfill the request.
'NULL:0/NULL:0/NULL:0 (state:STOCKOUT, sub-state:STOCKOUT, resource type:compute)'

A GPU stockout is routine, has nothing to do with your account, quota or billing,
and retrying in the same zone does not help. It had been surfacing downstream as
failed to connect to port 22, which sent us looking at firewalls, IAP roles and
SSH keys — none of which were ever wrong.

Now:

Google has no L4 capacity in us-central1-a right now, so comfy-win cannot start.
This is not a fault on your side and retrying in the same zone will not help.
to fix: wait and try later, or move the box to another zone.

ComfyUI was installed, but had no venv

.\venv\Scripts\python.exe : The term '.\venv\Scripts\python.exe' is not recognized

The launch hardcoded a venv path. A ComfyUI install may carry a venv, the Windows
portable bundle's embedded python_embeded, or neither. The interpreter is now
discovered on the box, the chosen path is printed — otherwise a wrong-Python
failure reads as a ComfyUI failure — and NO_PYTHON is reported plainly when there
is nothing to run at all.

Still untested live

The launch itself. us-central1-a has no L4 capacity at the moment, so the box
cannot be started to try it. The discovery covers the layout the box actually
reported, but first contact may still find something.

216 tests, 8 new.

🤖 Generated with Claude Code

https://claude.ai/code/session_015X8c3zczxydWiPfhnxfYNx

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/Comfy-Org/comfy-qa-tools/pull/17 **Author:** [@aliranjah](https://github.com/aliranjah) **Created:** 8/25/2026 **Status:** ✅ Merged **Merged:** 8/25/2026 **Merged by:** [@aliranjah](https://github.com/aliranjah) **Base:** `main` ← **Head:** `fix/stockout-and-python` --- ### 📝 Commits (1) - [`ff3f014`](https://github.com/Comfy-Org/comfy-qa-tools/commit/ff3f014dfe0fc956a1727485d0e97c74765e6b4d) fix: name a capacity stockout, and find Python instead of assuming it ### 📊 Changes **6 files changed** (+128 additions, -4 deletions) <details> <summary>View changed files</summary> 📝 `comfy_qa/lifecycle.py` (+25 -0) 📝 `comfy_qa/provision.py` (+31 -4) 📝 `docs/troubleshooting.md` (+11 -0) 📝 `tests/test_docs.py` (+2 -0) 📝 `tests/test_lifecycle.py` (+38 -0) 📝 `tests/test_provision.py` (+21 -0) </details> ### 📄 Description Two findings from a live run, one of which had been misleading us for an hour. ## The box could not start — Google had no L4 capacity ``` The zone does not have enough resources available to fulfill the request. 'NULL:0/NULL:0/NULL:0 (state:STOCKOUT, sub-state:STOCKOUT, resource type:compute)' ``` A GPU stockout is routine, has nothing to do with your account, quota or billing, and **retrying in the same zone does not help**. It had been surfacing downstream as `failed to connect to port 22`, which sent us looking at firewalls, IAP roles and SSH keys — none of which were ever wrong. Now: ``` Google has no L4 capacity in us-central1-a right now, so comfy-win cannot start. This is not a fault on your side and retrying in the same zone will not help. to fix: wait and try later, or move the box to another zone. ``` ## ComfyUI was installed, but had no venv ``` .\venv\Scripts\python.exe : The term '.\venv\Scripts\python.exe' is not recognized ``` The launch hardcoded a venv path. A ComfyUI install may carry a venv, the Windows portable bundle's embedded `python_embeded`, or neither. The interpreter is now **discovered on the box**, the chosen path is printed — otherwise a wrong-Python failure reads as a ComfyUI failure — and `NO_PYTHON` is reported plainly when there is nothing to run at all. ## Still untested live The launch itself. `us-central1-a` has no L4 capacity at the moment, so the box cannot be started to try it. The discovery covers the layout the box actually reported, but first contact may still find something. 216 tests, 8 new. 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_015X8c3zczxydWiPfhnxfYNx --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi added the

pull-request

label

2026-10-31 22:02:41 +00:00

bsakaguchi referenced this issue

2026-10-31 22:02:41 +00:00

[PR #16] [MERGED] fix: go hid the real failure and then tried SSH against a stopped box #17

bsakaguchi closed this issue

2026-10-31 22:02:41 +00:00